Occasional papers of the Museum of Natural History, the University of Kansas

HARVARD UNIVERSITY

Library of the

Museum of

Comparative Zoology

OCCASIONAL PAPERS

APR 1 0 1989

of the ^D

MUSEUM OF NATURAL HISTORl?^^
The University of Kansas
Lawrence, Kansas

NUMBER 128, PAGES 1-25 MARCH 30,1989

PHYLOGENETIC RELATIONSHIPS

OF SEVERAL SUBFOSSIL ANSERIFORMES

OF NEW ZEALAND

By

Bradley C. Livezey'

Three genera of waterfowl (Anseriformes) are known only from subfossil
remains from New Zealand (Oliver 1955, Howard 1964): Cnemiornis Owen
1866, Euryanas Oliver 1930, and Pachyanas Oliver 1955. The first two are
represented by numerous well-preserved elements, and Cnemiornis is re-
nowned for its radical morphological modifications related to flightlessness.
Pachyanas chathamica is represented by relatively few skeletal elements
(Oliver, 1955) and is not discussed here; the genus currently is under study
independently by R. J. Scarlett (pers. comm.) and P. R. Millener (pers.
comm.). With the exception oi Cnemiornis, which has a comparatively long
history of taxonomic reclassification and description, these subfossil endem-
ics have received little attention from avian systematists since their original
description.

A phylogenetic analysis of Recent anseriform genera (Livezey, 1986),
based largely on comparative osteology, permitted a reappraisal of the
relationships and classification of these endemic waterfowl. In this paper I:
(1) present analyses of characters of Cnemiornis, Euryanas, and the extinct
New Zealand swan Cygnus sumnerensis; (2) construct phylogenetic trees for
these groups based on these characters; (3) propose a revised classification of
these taxa; and (4) discuss selected evolutionary and biogeographic implica-
tions of these findings.

'Museum of Natural History, The University of Kansas, Lawrence, Kansas 66045 - 2454
U.S.A.

2 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

MATERIALS AND METHODS

I examined specimens oiCnemiornis calcitrans and C. gracilis (= septen-
trionalis) at the British Museum (South Kensington), Canterbury Museum
(Christchurch), Otago Museum (Dunedin), and the National Museum of New
Zealand (Wellington). Elements of Cnemiornis illustrated herein were bor-
rowed from the British Museum (BM). Euryanasfinschi and Cygnus sumner-
ensis were studied using specimens held at the Otago Museum. I endeavored
unsuccessfully to borrow specimens oiEuryanas for purposes of illustration
and confirmation of characters; therefore the analysis of this genus must
remain preliminary. Skeletal specimensofanhimids,An5eA-ana5',Den^wc}'^na,
Cereopsis, Branta, and Chenonetta were made available by the Museum of
Natural History, University of Kansas (KU) and U. S. National Museum of
Natural History (USNM).

Fundamentals of phylogenetic (cladistic) analysis are detailed in Wiley
(1981). Characters of available elements were coded as described in Livezey
(1986), which in turn was based in large part on the comparative osteology of
Woolfenden (1961). Anatomical nomenclature follows Howard (1929),
Woolfenden (1961), and Livezey (1986). Characters are discussed in the
"phylogenetic order" proposed by Livezey (1986), i.e., from the most
inclusive (ordinal) characters to the least inclusive characters (those support-
ing subfamilial and tribal relationships); characters are coded as in Livezey
(1986). Most skeletal elements of the Recent genera relevant to this study
were illustrated by Livezey and Martin (1988). Phylogenetic trees were
generated using the PAUP program (Swofford, 1985) based on the criterion
of maximal parsimony of character change (Wiley, 1981).

ACCOUNTS OF GENERA

Cnemiornis spp.

Taxonomic history. Owen (1866) described and illustrated a variety of
postcranial skeletal elements of a previously unknown large, flightless bird,
which he named Cnemiornis calcitrans, from deposits in a limestone fissure
atTimaru, South Island. Owen (1866) compared the elements with those of
the moas (Dinomithidac) and the flightless gruiform Aptornis, but proposed
no systematic placement for the species. Hector (1873a, b, 1874) examined
more material for the species, including a skull and a complete sternum, and
recognized it as a member of the "Lamellirostrate Natatores" (= Anseri-
formes). Owen (1875, reprinted in 1879) confirmed this classification and,
based on the additional material collected since his earlier work, concluded
that the humerus attributed to Cnemiornis in the original description (Owen,
1866) was actually that of the flightless gmKorm Aptornis. Owen (1875) also

SUBFOSSIL ANSERIFORMES 3

presented detailed osteological comparisons of Cnemiomis with the modem
Cape Barren Goose (Cereopsis novaehollandiae) of Australia and a flightless
steamer-duck {Tachyeres cf. pteneres) of South America, and, finding that
Cnemiomis was more similar to the former, inferred that Cnemiornis was of
anserine affinity. This decision was to influence profoundly the subsequent
classifications of the genus.

Forbes (1890) discovered that the coracoids illustrated by Owen (1875,
1879) as those of Cnemiornis were instead those of the unique flightless
gruiform Aptornis, and stated that the coracoid of Cnemiornis closely re-
sembled that of Cereopsis. Forbes (1891) contributed to this perception of
close relationship with his report ofaCereopsis from New Zealand, to which,
on the basis of a subfossil cranial fragment, he gave the name novaezealan-
diae. Lydekker (1891) listed Cnemiornis calcitrans within the Cereopsinae,
and noted several small specimens that (p. 102) "...indicate a distinct species;"
he also illustrated a coracoid of Cnemiornis. Forbes (1891, 1892a, b) distin-
guished a smaller species of Cnemiornis of the North Island (C. gracilis) from
the larger South Island form (calcitrans); he (1891) also proposed a third
species, C. minor, on the basis of several tibiotarsi from the South Island.

Oliver (1930, 1945, 1955) followed Owen in his placement o( Cnemiornis
with Cereopsis, and he considered both to be "geese." Oliver justified this
classification using comparisons (mostly cranial) between the two genera, but
acknowledged a number of conspicuous dissimilariUes in osteology and did
not discuss comparisons with any other genera. Oliver (1930, 1955) also
proposed the name C. septentrionalis for the North Island form. Lambrecht
(1933) listed three species {calcitrans, gracilis, minor) of Cnemiornis, also
under the subfamily Cereopsinae, after the typical geese. Delacour (1954: 1 99)
endorsed this practice with the comment: "...the genus Cereopsis has no very
close living relative, although the exfinct Cnemiornis calcitrans from New
Zealand was probably similar."

Dawson (1958) re-evaluated several of the taxonomic decisions based on
the types designated by Forbes and found that: (1) the supposed Quaternary
record of Cereopsis "novaezealandiae" from New Zealand was based on the
misidentification of a fragmentary specimen of Cnemiornis calcitrans, and
(2) that C. septentrionalis Oliver is a junior synonym of C. gracilis Forbes.

Both Brodkorb (1964) and Howard (1964) adhered to the tradition of
lisUng Cnemiornis with Cereopsis but, in accordance with the re-assignment
of Cereopsis to the shelducks (Tadominae) by Delacour and Mayr (1945),
placed both genera in or next to this subfam ily. Delacour ( 1 964) dismissed the
osteologically based conclusion of Woolfenden (1961) and the ethological
inference of Johnsgard (1961a) that Cereopsis should be returned to the
Anserinae.

Character analysis. I examined virtually all major skeletal elements of
Cnemiornis, including the skull, humerus, radius, ulna, carpometacarpus.

4 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

femur, tibiotarsus, tarsometatarsus, sternum, coracoid, scapula, furcula, and
pelvis. Most were represented by several specimens. The quadrate, pterygoid-
palatine complex, and trachea were not available for study. The northern and
southern "species" of Cnemiornis differed in size but were identical in the
characters discussed below, and are considered together under the generic
taxon in descriptions and phylogenetic analyses. The skuU and a variety of
postcranial elements of Cnemiornis were illustrated in Owen (1866, 1875,
1879) and Hector (1893a, b).

Cnemiornis shared several derived characters with all others of the
Anseriformes, including bill lamellae (13b, indicated on the ventral surfaces
of the premaxillae), recurved and pointed retroarticular processes of the
mandible (14b), pedicellate basipterygoid processes (20b), and reduced
furcular process of the furcula (102b). Several synapomorphies support the
membership oiCnemiornis in the suborder Anseres (anseriforms exclusive of
the Anhimidae): occipital fontanelles are indicated in outline, although
(secondarily) closed, as they are in some specimens of large Anserinae (9b,
see fig. 3 in Hector 1873a); and the caudal terminus of the pubis shows some
ventral orientation (116b, although more sHght than in Anseranas; Fig. 1).

Several characters indicate that Cnemiornis represents a branch after
Anseranas (i.e., appears to be synapomorphic with the Anatidae, sensu
stricto). One of these is the rounded cranial terminus of the upper bill (12c).
Another is the lack of an iliac recess in the pelvis (120b, illustrated in plate
XIV of Hector 1893a). Two others are of questionable reliability because of
the radical morphological changes associated with flightlessness: the ap-
proximately equal distal extent of the facets for digits II and III of the
carpometacarpus (45b) and the absence of a pneumatic foramen in the dorsal
surface of the coracoid craniad to the sternal facet (93b). The third synapo-
morphy uniting Cnemiornis with the Anatidae is the absence of a facet for
metatarsal I on the caudal surface of the tarsometatarsus (71b).

A number of osteological characters show that Cnemiornis was more
primitive than Dendrocygna and the rest of the Anatidae, i.e., was symple-
siomorphic Wiih Anseranas and indicate that Cnemiornis diverged from other
Anseres prior to the Dendrocygnines. These characters include: the relatively
caudal orientation of the femoral head (51a); the approximately equal distal
extent of the tarsometatarsal trochleae for digits II and IV (68a; illustrated in
Owen, 1866, plate 67, and described by Owen, 1875, pp. 269-270), a
condition coincident with the lack of caudal rotation of the inner trochlea; a
moderate lateral displacement of the calcaneum on the tarsometatarsus (72a;
Fig. 2); the strictly cranio-caudal orientation of the (incompletely ossified)
distal foramen of the tarsometatarsus (77a; Fig. 3); the presence of a large,
densely margined foramen at the base of the procoracoid (92a, see below); the
long, wide, and rounded conformation of the stemocoracoidal process of the
coracoid (99a, shared also with Cereopsis); and the equal proximal extent of

SUBFOSSIL ANSERIFORMES

Fig. 1. Postaceubular portions of the pelvis, lateral surfaces: (A) Anseranas semipalmata
(KU 80620); (B) Cnemiornis calcitrans (BM 75.12.15.4); (C) Dendrocygna autumnalis (KU
37725); (D) Cereopsis novae hollandiae (US^fM 430244); and (E) Branta canadensis (KU
23403). Caudal margins of iUium and ischium (CM) and caudal tenminus of pubis (P) are
indicated on Anseranas.

OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Fig. 2. Proximal ends of the left tarsometatarsus. caudal surfaces: (A) Anseranas semipal-
mata (KU 80620); (B) Dendrocygna autumnalis (KU 37725); (C) Cnemiornis calcitrans (BM
75.12.15.32); (D) Cereopsis novae hollandiae (USNM 429730); and (E) Branta canadensis
(KU 23403). Calcaneum is shown in stipple.

the coracoidal process and acromion of the scapula (109a). Also, the costal
margin of the extremely modified sternum of Cnemiornis occupies slightly
less than half of the basin length (86a), a primitive proportionality found only
in anhimids and Anseranas among Recent anseriforms. Cnemiornis lacks,
however, all of the diagnostic autapomorphies oi Anseranas (Livezey, 1986).
Three subordinal characters (88, 100, 104) were problematic because of the
dubious homologies of states related to flightlessness; for example, the
reduced furcula of Cnemiornis shows moderate flattening of the clavicles,
reminiscent of that in the Anhimidae (104b), but the state was coded as
"missing" for Cnemiornis.

The presence of a conspicuous procoracoidal foramen (92a) in Cnemiornis
is especially compelling evidence of its primitiveness; this character typically
occurs among modem Anseriformes only in the Anhimidae and Anseranas
(Woolfenden, 1961; Livezey, 1986). This primitive character is variable in
conformational detail, however, and deserves more detailed description. A
densely margined foramen is characteristic of the anhimids (Chauna and

Fig. 3. Distal ends of the left tarsometatarsus, cranial surfaces: (A) Anseranas semipalmala
(KU 80620); (B) Dendrocygna autumnalis (KU 37725); (C) Cnemiornis calcitrans (BM
75.12.15.32); (D) Cereopsis novaehollandiae (USNM 429738); and (E) Branta canadensis
(KU 23403). DisUl foramen (DF) and trochlear groove (TG) are indicated; note also the relative
distal extent of lateral and medial trochleae in A through C vs. that in D and E.

SUBFOSSIL ANSERIFORMES 7

Anhima), Anseranas, and Cnemiornis (Fig. 4), although considerable vari-
ation occurs in at least Chauna (Livezey, 1986). A superficially similar
structure is found infrequently in Cereopsis (perhaps only in captive birds),
but is distinguishable (when present) by its thin medial margin, evidently an
ossified ligament. The "foramina" of most Cereopsis also differ from those
oi Anseranas and Cnemiornis in lacking an enclosed pneumatic foramen to
the interior of the element in their caudal margins. Variation of this character
in Cereopsis is indicated by the range of variation seen in a series of skeletons
of captive birds held in the U. S. National Museum (Fig. 4). This structure
varies considerably in Cereopsis, and differences occur even within individu-
als; the specimen with a completely closed foramen in its right coracoid (Fig.
4g) lacked the suggestion of closure in its left coracoid (i.e., resembled Fig.
4e).

Unfortunately, the early descriptions of the coracoid of Cnemiornis were
largely erroneous. Owen (1875, 1879) figured coracoids which he attributed
to Cnemiornis and Cereopsis, but those labelled as Cnemiornis are instead
those of the ^viiiovmAptornis (Forbes 1 890; pers. obs.), whereas the coracoid
attributed to Cereopsis also is assigned incorrectly. The latter (plate XXXVII
in Owen [1875]) is a coracoid with a prominent procoracoidal foramen and
appears to be identical to the coracoid of Cnemiornis and almost certainly
pertains to that genus. Lydekker (1891) illustrated a genuine coracoid of
Cnemiornis, but stated (p. 100) without supporting details that "This speci-
men (fig. 26) agrees very closely with the coracoid oi Cereopsis."" The notion
that a procoracoidal foramen occurs in some Anserinae was perpetuated by
the statement of Howard (1964:250) that "...this foramen is rarely found in the
Anatidae [sic] except in Anseranas and occasionally in certain swans."
Although no details were given, this observation by Howard probably
stemmed from the traditional assignment of certain primitive fossil anseri-
forms (e.g., Cygnopterus) to the Anserinae (Livezey, 1986) or to the infre-
quent foramen-like structures seen in procoracoidal processes of some
anserines (Fig. 4),

Several synapomorphies unite Cnemiornis with the Anatidae exclusive of
the Dendrocygninae: the caudal margins of the ilium and ischium present an
obhquely sloping aspect (114b; Fig. 1); the inner cnemial crest of the
tibiotarsus shows slight, perhaps equivocal lateral deflection (63b); and the
tarsometatarsal trochlea for digit II is grooved (74b, Fig. 3; see discussion of
homoplasy by Livezey and Martin, 1988).

Numerous characters of the wing and pectoral girdle, elements that were
modified substantially in association with the loss of flight in Cnemiornis,
were not comparable to the states defined for Recent anseriforms (Livezey,
1986). These characters, several of which present difficulties in comparisons
even among some flighted waterfowl, include features of the carpometacar-
pus (37, 38,43,44), sternum (78, 79, 81, 88, 89), coracoid (96, 100), furcula

8 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Fig. 4. Cranial portions of the left coracoid, ventral surfaces: (A) Anseranas semipalmata
(KU 80620); (B) Cnemiornis calcitrans (BM A.1521); (C) Dendrocygna autumnalis (KU
37725); (D) Branla canadensis (KU 23403); (E-G) Cereopsis novaehollandiae GJSNM
420244, 429738, 3 1 8044, respectively). Procoracoidal foramen (PF) and pneumatic area under
brachial tuberosity (BP) are indicated.

(101, 105), and scapula (108, 112).

Cnemiornis lacks the diagnostic synapomorphies of the true geese
(Anserini): it evidently retained the primitive number of 17 cervical verte-
brae, including the axis and atlas (21a; cf. partial counts in Hector [1873a, b],
reconstruction in Owen [1875], and mounted skeleton BM 7512154); it
shows no spur-like elaboration of metacarpal I (42a); there are no pneumatic
foramina under the brachial tuberosity of the coracoid (95a; Fig. 1); and its
pubes lack caudal flanges (117a; Fig. 1). Cnemiornis does not share the
autapomorphic supraorbital process (11a), pneumatic swelHng of the fronto-

SUBFOSSIL ANSERIFORMES 9

nasal region (16a), or dorsal bowing of the upper bill (19a) characteristic of
Cereopsis.

Neither Brodkorb (1964) nor Howard (1964) offered ostcological support
for their placement of Cnemiornis with the shelducks (Tadominae); this
assignment evidently resulted from its traditional association with Cereopsis,
itself not tadomine (Woolfenden, 1961; Livezey, 1986). However, it seems
prudent to review the skeletal evidence against a close relationship between
Cnemiornis and the Tadominae. In addition to the symplesiomorphies of
Cnemiornis and Anseranas discussed above, several character states of
Cnemiornis are primitive relative to those of the larger clade including
Stictonetta, Plectropterus, the Tadominae and the Anatinae: the retention of
the primitive number of cervical vertebrae (21a); the orientation of the
humeral capital shaft ridge toward the head (22a); the short capital groove of
the humems (23a); the unelevated humeral facet of the anterior articular
ligament (26a); the proximally rotated intemal tuberosity of the humerus
(27a); the dorsal surface of metacarpal n is flattened proximally (39a); the
attachment site of M. extensor metacarpi ulnaris on the carpomelacarpus is
completely proximad to the proximal fomix (43a); and the lack of a medial
protuberance in the ventral manubrial region of the stemum (79c). In addition,
Cnemiornis is more primitive than the Tadominae sensu stricto in the
unenlarged process of metacarpal I (42a) and the tibiotarsus without torsion
about its long axis (61a).

Euryanas finschi

Taxonomic history. Van Beneden (1875) described a small duck from
subfossil remains found in Eamscleugh Cave, New Zealand, and named it
Anas finschi. This paper, in French, was followed by a report in English (Van
Beneden, 1876). In both papers. Van Beneden compared elements oi finschi
variously with those of the Recent Dendrocygna eytoni. Anas gibberifrons,
Aythya fuligula, and Bucephala clangula, as well as to the Miocene fossil
Mionetta (^"Anas") blanchardi (Livezey and Martin, 1988). Hamilton (1892)
reported the discovery of more specimens oi finschi in the fissures at Castle
Rocks.

Oliver (1930) placed the species in its own genus Euryanas, which he
(1930, 1945, 1955) believed compared favorably (using skull characters)
with the Maned Duck (Chenonettajubata), a morphologically and behavior-
ally unique endemic of Austrialia (Delacour, 1959). Lambrecht (1933) placed
E. finschi within the Anatinae. Howard (1964), following the assignment of
Chenonetta to the "perching ducks" ("Tribe Cairinini") by Delacour and
Mayr (1945) and Delacour (1956), placed Euryanas in this tribe. Without
comment, however, Brodkorb (1964) listed Euryanas within the "spur-
winged geese" (his Plectropterinae).

10 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Character analysis. I examined specimens of the skull (lacking quadrates
and pterygoid-palatine complex), humerus, ulna, carpometacarpus, femur,
tibiotarsus, tarsometatarsus, sternum, costae, coraroid, furcula, scapula, and
pelvis of Euryanas. S. L. Olson {in litt) described the syringeal bulla of
Euryanas as being of typical anatine form (sensuAnas, Chenonetta; 6c). The
skull and a number of the postcranial elements of Euryanas finschi were
figured by Van Beneden (1875, 1876).

Inclusion of Euryanas within the suborder Anseres is supported by all
available characters listed by Livezey (1986), including its typically "duck-
like" bill. Synapomorphies uniting Euryanas with other Anatidae (sensu
stricto, excluding Anseranas) are equally numerous , including: the craniome-
dial orientation of the femoral head (51b), the lateral deflection of the inner
cnemial crest of the tibiotarsus (63b), the proximal position of the tarsometa-
tarsal trochlea for digit II (68b), the orientation of the distal foramen of the
tarsometatarsus (77b), the absence of a procoracoidal foramen (92b) or an
iliac recess in the pelvis (120b). Two synapomorphies show Euryanas to be
derived with respect ioDendrocygna — a grooved tarsometatarsal trochlea for
digit II (74b) and the obliquely sloping caudal margins of the ilium and
ischium (114b).

Additional synapomorphies support a closer relationship of Euryanas
with Stictonetta + Plectropterus + Tadominae + Anatinae than with Thalas-
sornis or the Anserinae: the orientation of the capital shaft ridge (22b) and
capital groove (23b) of the humerus; and the presence of a notch (although
weak) in the external rim of the carpal trochlea (38b), the rounded dorsum of
metacarpal II (39b), and the position of the scar of M. extensor metacarpi
ulnaris (43b) of the carpometacarpus. With the exception of a widening of the
scapular blade (108b), Euryanas lacks the synapomorphies characteristic of
the Anserinae (e.g., characters 85a, 95a; Livezey, 1986).

Compared with the shelducks, Euryanas is primitive in the unelevated
facet for the anterior articular ligament (26a) and proximally oriented internal
tuberosity (27a) of the humerus, and lacks the tadomine synapomorphies of
a carpometacarpal spur (42a) and tibiotarsal torsion (61a). Euryanas is
plesiomorphic with respect to the Anatinae in a number of characters, notably
in the rounded, anconally concave deltoid crest (25a) and prominent, but-
tressed external tuberosity (32a) of its humerus.

These characters indicate that Euryanas diverged from modern anatid
lineages after the basal anatid grade of Dendrocygna, Thalassornis, and the
Anserinae, but before the Tadominae (Livezey, 1986). Euryanas lacks the
somewhat convergent features indicative of diving specialization found in
Thalassornis, pochards (Aythyini), sea ducks (Mergini), and stiff-tailed
ducks (Oxyurini), especially characters of the femur (52a, 54a, 55a, 56a),
tibiotarsus (64a, 65a), tarsometatarsus (69a, 75a), sternum (78a), and pelvis
(119a).

SUB FOSSIL ANSERIFORMES 1 1

Both Euryanas and Stictonetta have long, peg-like ventral manubrial
spines (79d), a character shared also by the more derived genus Anas; this
feature, however, is variable and its transformational pattern is inadequately
resolved (Livezey, \9S6). Euryanas differs from Stictonetta in two characters
of the coracoid, both of problematic polarity and transformation: the ventral
surface is without a deep depression (96b) and the ventral sternal facet is
without a buttress (100a).

The presence of an asymmetrically enlarged, unfenestrated syringeal bulla
in Euryanas (6a; S. L. Olson, in litt.) supports, however, a closer relationship
between Euryanas and the terminal clade of Plectropterus + Tadorninae +
Anatinae than with Stictonetta.

Cygnus sumnerensis

Taxonomic history. Forbes (1890a, b) and Sclater (1890) announced the
discovery of three coracoids and a partial humerus of a large, extinct swan
from a cave near Christchurch, which Forbes (1890a) named Chenopis
sumnerensis. Forbes (1891) reported the collection of more material for the
species, and speculated that more than one species might be represented.
Forbes (1893a, b) later reported numerous specimens of the swan from the
Chatham Islands.

Oliver (1930) and Lambrecht (1933) listed Chenopis sumnerensis as a
typical swan. Oliver (1955) later reassigned the fossil swan to Cygnus in
accordance with current generic taxonomy; he also proposed a new species
name, C. chathamicus, arguing that the earlier name should be abandoned
because the types for the species described by Forbes ( 1 890a, b) could not be
identified. Dawson (1958) reported the rediscovery of this type material and
relegated chathamicus to junior synonomy of C. sumnerensis, a restoration
followed by Brodkorb (1964) and Howard (1964).

Character analysis. I examined all important skeletal elements of C.
sumnerensis except the quadrate, pterygoid-palatine complex, and syrinx.
Oliver (1955:603) figured a mounted, presumably composite skeleton of this
species.

C. sumnerensis is synapomorphic with modem geese and swans (Anseri-
nae) in the presence of foramina on the midline and cranial margin of the
dorsal surface of the sternal basin (89a), the presence of foramina under the
brachial tuberosity of the coracoid (95b-c), the lack of a ventral depression
on the coracoid (96b) , reduced coracoidal tuberosities on the furcula (101 a-b),
and a caudal widening of the pubis (1 17b). The species is united with extant
swans (Cygnini) by the caudomedial extension of the xiphial region of the
sternal basin (85b), the presence of a small foramen in the cranial edge of its
uninflated sternal carina (87b), and ±e comparatively medial orientation of
the sternal intermuscular line (88a). Two apparent synapomorphies are shared

12 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

with the modem geese (Anserini): an enlarged process of metacarpal 1 (42c)
and the diverse foramina present under the brachial tuberosity of the coracoid
(95c). The moderately low consistency and sexual variation of the first
(Livezey, 1986), the intermediate condition of the second, and the hmited
material available for characterization of C. sumnerensis support the interpre-
tation that these similarities are convergent. C. sumnerensis lacks the derived,
trachea-related modifications of the sternal carina and furcula found in Olor
(87c, 106b).

CONSTRUCTION OF TREES

Methodological Considerations

For derivation of phylogenetic trees, I used the characters described in
Livezey (1986), but excluded from analyses those characters which were not
informative for inferences concerning relationships among subfamilies.
Excluded characters were unique autapomorphies (particularly of Anhimidae,
Anseranas, and Plectropterus), diwing-rthted autapomorphies oiThalassor-
nis (several convergent with some members of the Tadominae and Anatinae),
and characters which were invariant among the Anseriformes exclusive of the
Tadominae and Anatinae. This reduced the characters analyzed to 62 which
were useful for inferences in the basal segment of the order (Figs, 2 and part
of Fig. 3 in Livezey, 1986), the segment which, on the basis of the foregoing
character analyses, included the subfossil genera to be placed. As in Livezey
(1986), several characters were analyzed as unordered (Table 1). In all
analyses, two primary weighting schemes were employed: the "standard"
weighting scheme of Livezey (1986), in which the syringeal bulla (character
6) was given a weight of two and all other characters were given unit weight;
and the "unit" weighting scheme in which all characters were given unit
weight.

A further simplification was made through the reduction of the taxonomic
units considered in construction of trees. Recent taxa analyzed were reduced
to 12 taxonomic units, in addition to the hypothetical ancestor proposed by
Livezey (1986): the Anhimidae, seven single-genus lineages {Anseranas,
Cereopsis, Coscoroba, Dendrocygna, Thalassornis, Stictonetta, and Plec-
tropterus), and four taxa representing well-established monophyletic groups
of genera {Branta-Anser, Cygnus-Olor, Tadominae, and Anatinae). This
streamlined set of Recent taxa was used for separate (14-taxon, 62-character)
phylogenetic analyses of Cnemiornis and Euryanas using the exhaustive
branch-and-bound algorithm in the PAUP program, a time-consumptive
technique for finding all possible shortest trees which is practical only with
small numbersof taxa (Swofford, 1985). The body of evidence confirming the
systematic position of Cygnus sumnerensis among the moderately derived

SUBFOSSIL ANSERIFORMES

13

« o

oa-a

C rt

^ £

o <u

^s

<u o

-<-• u.

2 (X

'^ ,

T3 ' .

C X

C3 -

£ C

dj a

X) c/,

S Ci

3 C

= ?

u. o

1/5
•1— >

O C/5

<u

2 2i

^

o

c3 ca

03

J=

s-

U 2 ^

«

>-,

^ -o

n

^

S.s

C/3

"*- C/3

c

<r3

t4- C

O w

« C/5

oh

1- ^

c

gen
steri

c^

rz: «

"*

^:S

C/!)

cd

■§1

T3

D

C/3 Ca

-4— t

03

B-g

-4—'

■ -^

^

■§ 2i

1>

c3 eg

>
4—)

?5 "^

C3

t*r, IJ

C

bXj V-

^

03 aj

<U

4-*

13

c

C 3

.^o

C3 c«

5

o

X «

<u -o

^

_, IJ

•^

W N

^

V) ><

^

^1

1- «

o

X fo
ters

1/3

'^ ':i

C3

^ cS

03 u.

V3

£^

<u U

■«-»

2 ^

c/3 (o

u. 00

(U 0^

4—1 ,— ,

o -^

03

V- >^

w aj

£ N

U ^

^ J

3 ^

E^£

U

ca

CS

^.^

i^

ca

00

03

C3

I <— \

=.

\D

in

*

'^

•X-
(N

'^

ON

m

oo
m

m
(^

r--

<N

SO
(N

m

CN

*

(N

O
(N

ON

00

so

*

O

ON

*

(N

c

03
03

03

CC

03
03
03
03
eg
03
03

03
C«
eg

eg

03

eg

03
03
eg

03
03
03

c/:
(L)

c
<

eg eg eg^X)XiX)XiX)XiX!X)
eg cgj2X)XiX)X)XiX)X)X3^

03 03X)Xi^X)-OXiXiXi^X3
eg egXiXJXiXiXiXJjnXiX)^
eg egXiXiXi^X)X)Xi^X)X)
ego3030303cgo30303cgX)o3

eg egX)^X)^X)X)XijDXi^

03 cgXiX3X)XiX) cgX)X)^X)

eg eg =^^X)X)-OJ2jD£iX)^
egcgc^cgog-Oegog^cg-D^

_CCgWo300cgo3o3o3003
o3o30303cgcgo3c«XiX>X!^

eg eg 03 c«X)X)XlXiX)Xl^^
ego3o3o3o3ego3cgX>cgc3Xi
cgo3cgX)o3o3c3cgcgo3cgXi
o3o3o3X)o3cgegego3cgX)X)

O303030030303cgcao3^-^
ego303o3cgo30303O3O3O3O3

egcgcgX>o3cgo3cgXiXi^X5
03o3cgo3o303cgcgXiXiXlO

o3ego3o3X)-dO"Oo3cgu!U

o3cgo3o3Xio3o3o3cgcgo3o3
egX)X)X)XiX)X)-OXiXiX>Xi
(Tiriegegejcgegegegcgego3
XiX)XiXiego3o3cgX)cgo303

egXioooooocjooo

03cgo3o3_D0303cgo3X>Xio3

X>o3-oo(uu<u«"uajiuoj

cgjOX>-C3Xl^XiXiX)^X)Xi
0303Xlc003030303cgejOO
cgcgcgo3ego30303C3Xi-OX5

~ -2

^

^

£i

Xi

03

X

X)

Xi

eg

X>

X)

Xi

03

eg

eg

X3

Xi

X3

X

X

"m"

X)

eg

03

eg

Xi

'w'

Xi

X

X

X

eg

03

eg

03

eg

X

eg

03

Xi

eg

Xi

eg

X

X3

X)

eg

X

X

X

03

X

X

X

JD

X)

J3

X)

u

a

03 X

X <U

X u

X X

.in

.2 2

14 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

-a

3

c

c
o
U

C3

u

o

1—1

ON

O

00

o

o

O

*

o

ON
ON

*

ON
*

ON

rri

ON

CM

ON

*

ON
00

OO
OO

r-

00

m

00

*

ON

*

OO

c

s

f2

«

03

C3

«

CO
CO

X

c«
03
03
X
03

X
X
X

c

<

03 03X)XiX).£iX5J2X)^-O^ XiX3

«03X!-OX)X)X)X)X)Xl^X2 Is -O

0303C3P3XlXiXiXio3o3o303 C003

03XiX)X)XJX).OX)XiX)J2X) XiXi

c3cao503o303X)X)c0o3oSc3 XC3

c3C3 03X)X)^X>^XJX3X1X) X!^

c3cacac0X)W030303c003cg XX

o303X)X>cO:OWXi^X)X>Xi XX)

Ci5 03X)X)XiX)^XlX>XlX)X) WX)

03O3o3COXiX)XiXJO303o3CO XX5

03c00303Xlc0X3O03030303 0303

^C3COCOCOCOCOC3C003COCO

Xi OXiX)^X)XXiX3XiX)X)
X)X) 03 cO^X)XiX3 03X) 03 c3

«X>-OX)XiX)X)X)X)X) « 03
CO C3XX) 03X)X^X)XXiX)
XJXl CO cOXiXlXl^ 03^^X1
03C0C003OO.OX5C3C00303

CO 03X)X)X)XiX)XlXiX)XlXi
03 cOXiX)XiX)XiXXiXiXiX
CO oXJX) CO CO CO cOXlXiXlX)

OX)X>X)X>X) CO 03X)XiXiX)

COCOCOCOCOCOX|X5C00303CO
C0030303COCOX)X)03C003CO

cOC0cO03o3c0C003XlXiXiX)

XiXiOCJOCOcOCOTSOiUU

r^ X) COCJcOCOcOCOCOCOCOCO

03 COXiXiXX^X)XX)XlX)

'^ 1^ ^

'^ C^ jS

^ &.•= CO

X) X)

X CO

'^ 03

03 Xi

X XI

CO CO

X) X)

CO X)

X X)

X Xi

X 03

X CO

X X

2 "^

X 03

CO x:>

•2 c

SUBFOSSIL ANSERIFORMES 15

members of the modem genus Cygnus rendered unnecessary a numerical
analysis of the species; a species-level analysis of Cygnus is not possible at
present.

Cnemiornis spp.

The phylogenetic relationships of this genus were analyzed under a variety
of alternative character codings and weights. An initial analysis was based on
all 62 characters, for which 41 were determined for Cnemiornis (Table 1); all
other character states were coded as "missing." A series of subsequent,
progressively "conservative" analyses were performed in which an additional
1-7 characters considered to be problematic for Cnemiornis — two of the
carpometacarpus (38, 43), one of the sternum (79), one of thecoracoid (100),
two of the furcula (104, 105), and one of the pelvis (118) were coded as
"missing" as well. Both the "standard" and "unit" weighting schemes were
employed. In addition, each of these analyses was performed with another
somewhat problematic character, the procoracoidal foramen (92), assigned a
unit weight Gike all other characters) or with weight zero (i.e., it did not affect
the derivation of trees); the latter treatment seems justified in light of
intrageneric variation in this character and the possibility that the foramen
typical of Cnemiornis is homologous to those found infrequently in Cereopsis
(Fig. 4). None of these analytical variants altered the position of Cnemiornis
in the resultant trees or the number of equally short trees found, but these
modifications of data did produce minor differences in tree lengths and
consistency indices. ConsequenUy, the trees depicted in Fig. 5 are based on
the most conservative analysis of Cnemiornis, in which 34 character states
were specified for the genus.

In all analyses, Cnemiornis was inferred to be a lineage which diverged
after Anseranas but before the divergence of the Recent taxa included in the
Anatidae by Livezey (1986); i.e., Cnemiornis was found to be the sister-group
of the suborder Anseres exclusive of Anseranas (Fig. 5). The position of
Cnemiornis was supported by 15 character changes, between Anseranas and
Cnemiornis, and seven character changes supportive of monophyly of the
other Anseres. Two characters which were retained in all analyses and which
were derived in Cnemiornis but primitive in Dendrocygna (characters 74,
114) were inferred to be reversals in the latter.

An unexpected finding of the branch-and-bound analyses of the reduced
data set for extant taxa was that several, equally parsimonious permutations
oi Dendrocygna, Thalassornis, and Ihe Anserinae are possible (Fig. 6). Only
one of these is most parsimonious if Cnemiornis is included in the analysis
(Fig. 5); this arrangement, in which the Anserinae are inferred to be the sister-
group to the rest of the (Recent) Anatidae, differs from both of the two
topologies described by Livezey (1986).

16 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Ancestor

Anhimidae

Anseranas

LENGTH = 106
CI = 0.76

Den droc ygna
— Thalassornis

St ictonetta

Plec tropterus
Tadorninoe
Ana 1 1 nae

-d

-c

C oscoroba

K

Cygnus + Olor
Br onto + Anser
Cereopsis

CNEMIORNIS

Cygnus + Olor

Cygnus + Olor

c

Coscoroba

Sranta + Anser

' — Cereopsis

• Branta -'r Anser

Fig. 5. Phylogenetic trees for Cnemiornis and basal lineages of Anserifonmes: (A) Complete
tree with one of three equally parsimonious tqjologies (for branches overlaid with stipple) for
the geese (Cereopsis, Anser, and Branta); (B, C) Alternative tq)ologies for the geese.

There also were three equally parsimonious topologies within the Anseri-
nae, alternatives which were possible whether Cnemiornis was included in
the analysis or not (Fig. 5). These topological variants are not directly relevant
to the position of Cnemiornis in the tree, but one of these topologies — in
which Cereopsis is hypothesized to be the sister-group to the rest of the
subfamily (Fig. 5b) — would be favored in the event that symplesiomorphies
between Cnemiornis and Cereopsis (not shared by other anserines) were to be
documented. A strict consensus tree for the analysis of Cnemiornis depicts the
Anserinae as a trichotomy involving Cereopsis, Branta + Anser, and the
Cygnini.

Euryanas finschi

Of the 62 characters employed in the analyses of subfossil taxa, 50 were
determined [or Euryanas (Table 1); the other 12 were coded as "missing." An
exhaustive search for all most-parsimonious trees using the "standard"
weighting scheme found nine equally parsimonious topologies, but in each
Euryanas was the sister-group to the clade Tadominae -i- Anatinae (Fig. 7a).

SUBFOSSIL ANSERIFORMES

17

Ancestor

A nhimidae

— An ser anas

LENGTH = 106
CI = 0.76

Dendrocygna
• Anser I noe

— Dendrocygna

T halassornis

T holossornis

other Anatidoe

■ Coscoroba
• Cygnus ■*- Olor
— C ereopsis

' Branta -t- Anser

St I ctonett a
Pleclropterus

CTadorn I nae
Anoti nae

' Anserinoe

Anser i n oe

Dendrocygna

- T halassornis

other Anotidae

Fig. 6. Phylogenetic trees for extant basal lineages of Anseriformes: (A) Complete tree with
oneof three equally parsimonious topologies (for branches overlaid with stipple) ior Dendrocygna,
Thalassornis, and Anserinae (the last taxon shows three topologies depicted in Fig. 5); (B-C)
Alternative topologies for the grade including Dendrocygna, Thalassornis, and Anserinae.

The topological variants resulted from two previously discussed, unresolved
segments: (a) three alternative positions oWendrocygnus, Thalassornis, and
Anserinae; and (b) the three arrangements of Cereopsis, Branta + Anser, and
the Cygnini within the Anserinae. Using the "unit" weighting scheme, 12
equally parsimonious trees were found, but once again Euryanas was the
sister-group to the clade composed of Tadominae and Anatinae in each. The
majority of the topological variants resulted from combinations of arrange-
ments within the unresolved grade and geese (discussed above); reduced
weight of the syringeal bulla, however, permitted an additional sequence for
the grade composed of Sdctonetta and Pleclropterus (three of 12 trees; Fig.
7b).

PROPOSED CLASSIFICATION

Based on the trees discussed above, I conclude that:

(1) Cnemiornis is the sister-group to the Anatidae (sensu Livezey, 1986).

(2) Inclusion of Cnemiornis in the phylogenetic analysis indicated that the
subfamily Anserinae (true geese and swans, including Cereopsis) may be the
sister-group to the rest of the family Anatidae (including Dendrocygninae).

(3) Euryanas appears to be a moderately derived "proto-duck," a member
of a lineage that arose after the Anserinae, Dendrocygninae, and Thalassor-
nithinae but before the divergence of the Tadorninae from the Anatinae; it
appears to be the sister-group to the clade composed of the Tadominae and
Anatinae.

18

OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Ancestor

Anhim idae

— Anser a nas

B

(— Plectropterus

Stictonetta

Tad or ni nae
Anat i nae
EURYANAS

LENGTH =110
CI = 0.74

Dendrocygna
I — T halassornis

Anserinae

I — St ic tonetta

Plectropterus

Tadorninae
Anati nae
' — EURYANAS

Fig. 7. Phylogenetic trees for Euryanas and basal lineages of Anserifonmes: (A) Complete
tree using weighting scheme of Livezey (1986), unresolved grade overlaid in stipple; (B)
Alternative topology for terminal clade involving Euryanas if syringeal buUa (character 6) is
given unit weight.

(4) Cygnus sumnerensis is a true swan (Cygnini), more derived than
Coscoroba but less so than Olor; it agrees in its characters with those of the
possibly paraphyletic genus Cygnus {sensu stricto; Livezey, 1986), which
includes C. atratus of Australia.

Accordingly, I propose the following classification, based on the relevant
section of the schemes presented by Livezey (1986) and Livezey and Martin
(1988), and annotational conventions of Wiley (1981). Taxa analyzed herein
are shown in boldface. Sedis mutabilis indicates that the phylogenetic
sequences of associated sections are unresolved.

Order Anseriformes
Suborder Anseres

Family Anseranatidae
Family Cnemiornithidae Stejneger, 1885
Genus Cnemiornis Owen, 1865
C. calcitrans Owen, 1865
C. gracilis Forbes, 1891
Family Anatidae

Subfamily Dendrocygninae^
Subfamily Dendrocheninae^
Subfamily Thalassomithinae'
Subfamily Anserinae^

Tribe Anserini (possibly paraphyletic)
Tribe Cygnini

^sedis mutabilis

SUBFOSSIL ANSERIFORMES 19

Genus Coscoroba

Genus Cygnus (possibly paraphyletic)
C. sumnerensis (Forbes, 1890)

Genus Olor
Subfamily Stictonettinae

Genus Stictonetta
Subfamily Plectropterinae
Subfamily Euryanatinae, subfam. nov.

Genus Euryanas Oliver, 1930

E.finschi (Van Beneden, 1875)
Subfamily Tadominae
Subfamily Anatinae

Diagnoses for Cnemiomithidae and Euryanatinae are as for the included
genera (see character analyses above). The Cnemiomithidae can be charac-
terized additionally by large size and reduced pectoral elements, both of
which were illustrated and described previously (Owen, 1866, 1875; Hector,
1873a, b; Howard, 1964); representative measurements were given by
Howard (1964). Note that in the foregoing partial classification the subfamily
name Thalassomithinae is used instead of the incorrectly derived taxon
Thalassominae given in Livezey (1986).

DISCUSSION

This reappraisal indicates that Cnemiornis and Euryanas represent two
variably primitive lineages endemic to New Zealand. The earlier classifica-
tions of these taxa were based largely on comparisons that were influenced
profoundly by biogeographic preconceptions; classifications oi Cnemiornis,
Euryanas, and Cygnus sumnerensis were based largely on comparisons with
Cereopsis, Chenonetta, and Cygnus atratus of Australia, respectively. This
was a strangely parochial approach to the study of waterfowl, an ancient
group in which several modern genera have cosmopolitan distributions (e.g.,
Cygnus, Tadorna, Anas). The early systematic analyses of these endemics
also were limited by the taxa compared (e.g., Owen, 1875) and the virtual
exclusion of postcranial characters from the pioneering work of Oliver (1930,
1945, 1955). The failure of previous workers to distinguish between primitive
and derived characters undoubtedly contributed to these classificatory prob-
lems, as has been the case for many paleomithological investigations (Crac-
raft, 1980). The methodological and philosophical justification for cladistic
analysis in paleontological study was reviewed by Schoch (1986).

The early perception of an "alliance" between Cnemiornis and Cereopsis
evidently was based on very limited phenetic comparisons and the compara-
tively large body size of both genera. Cnemiornis was extremely derived

20 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

osteologically and of immense size (by anseriform standards); both charac-
teristics were related to its obvious flightlessness, and reductions of wing
elements and the sternal carina in this genus are among the most extreme in
the Anseriformes (Livezey, in prep.). However, in virtually all other charac-
ters the genus is very primitive. The unique skeletal characters of Cnemiornis
led Oliver (1945) to conclude the genus deserved subfamilial rank, while
paradoxically maintaining the view that Cnemiornis "is a close ally of
Cereopsis" (p. 125). The traditional view of such an "alliance" was so
profound that, following the re-classification of Cereopsis to the Tadominae
(largely on behavioral grounds) by Delacour and Mayr (1945), Cnemiornis
was similarly reclassified by Howard ( 1 964) and Brodkorb (1964). Ironically,
Howard (1964) herself had warned against such taxonomic revisions (in the
context of the classification of a fossil anserine), indicating (pp. 268-269)
"...the need for caution in attempting to trace an evolutionary line based on the
names of fossils without careful review of the characteristics of the fossils
themselves in the light of accumulating knowledge of existing forms."

Wetmore (1943) described the fragmentary tibiotarsus of a goose from
Hawaii, Geochen rhuax, which he felt resembled Cereopsis (and by associa-
tion of traditional taxonomy, Cnemiornis) most closely. As with Cnemiornis,
Geochen followed the later movement of Cereopsis to the shelducks in both
Brodkorb (1964) and Howard (1964). In contrast, the present study indicates
that: (1) Cnemiornis is not a goose and is not closely related to Cereopsis (a
true goose, Anserini) but instead represents a very early branch of the
Anseriformes (i.e., is the sister-group to the Anatidae, sensu stricto); and (2)
neither Cnemiornis nor Cereopsis belongs in the shelducks (Tadominae). The
finding that Cereopsis is anserine and not tadomine in relationship was
inferred by several earlier investigations of osteology (Shufeldt, 1913;
Verheyen, 1953; Woolfenden, 1961; Livezey, 1986). Furthermore, prelimi-
nary examinations of Geochen and flightless Thambetochen of Hawaii
(Olson and Wetmore, 1976; Olson and James, 1982) indicate that these genera
are probably anserine and not tadomine in relationship.

It has been suggested to me that several of the "primitive," non-gooselike
characters found in Cnemiornis (especially of the pelvic limb) may be
reversals associated with the terrestrial specialization, a condition related in
tum to the evident flightlessness of the genus. Unfortunately, no compelhng
evidence for such functional relationships or for the occurence of such
reversals in other anseriforms has been demonstrated. Lacking such support,
and given that several other "terrestrial" waterfowl (e.g., Thambetochen,
Branta sandvicensis, Chloephaga) show none of these putative reversals (cf.
Miller, 1937), the parsimonious inference remains that these characters
reflect an early divergence for Cnemiornis. "Adaptive" rationalizations to
retain Cnemiornis in the Anserini, and in particular to suggest that Cnemiornis
is the sister-group of Cereopsis, not only assume synapomorphies not in

SUBFOSSIL ANSERIFORMES 21

evidence, but also imply additional homoplasy in the associated phylogenetic
hypothesis and are based on unsupported ad hoc arguments concerning
presumed evolutionary change. A similar suite of "adaptational" rationaliza-
tions was suggested by Davies and Frith (1964) to conserve the classification
of Anseranas within the Anserinae in the face of growing morphological
evidence of its extreme primitiveness(cf.Delacoiir, 1954;Johnsgard, 1961b;
Woolfenden, 1961).

Based on my studies of the osteology of waterfowl and the limited
comparisons involving Cnemiornis, I suggest that the following structures
may prove useful for further testing of the systematic position of the genus:
the conformational details of the palate; patterns of cranial canals, including
the morphology of foramina in the cranioventral floor of the cranium; the fine
structure of vertebrae, especially the cervical vertebrae and those composing
the synsacrum; and anatomical details of the calcaneum of the tarsometatar-
sus. Furthermore, I predict that the determination of homologous states and
transformation series of such fine-grained characters may prove problematic
if attempted for the entire order Anseriformes; study of a more restricted
subgroup of taxa (e.g., Anseranas, Cnemiornis, Dendrocygna, Cereopsis,
Anser, and Stictonetta) probably would be adequate for inferences regarding
Cnemiornis. At least one of these character complexes — the morphology of
the vertebral column — would provide insights into the augmentation of
cervical vertebrae in Anseranas and the Anserinae and thereby also shed hght
on the relationships within the Anserinae.

Euryanasfinschi also deserves continued study. The unfortunately limited
material available for this study notwithstanding, there is substantial evidence
that Euryanas is not anatine or tadomine {sensu Livezey, 1986); this lends
some support to the observation by Oliver (1 945: 1 24) thatEMryano.y "...seem[s]
to be more primitive than the typical ducks." However, I am not persuaded by
Oliver (1945) that Chenonetta and the New Zealand teal {Anas chlorous and
A. aucklandica) are similarly plesiomorphic, although Chenonetta is osteol-
ogically unusual in several respects (Woolfenden, 1961; Livezey, 1986).
Skeletal elements deserving of particular attention in Euryanas are the skull,
carpometacarpus, and tibiotarsus; the (modal) number of cervical vertebrae
would be particularly useful for phylogenetic inference. Worthy (1988)
inferred that there has been modest shortening of wing elements in Euryanas
during recent millenia; whether this trend has modified any of the osteological
characters considered here is not known. Two humeral characters which I
found to be problematic in Euryanas (characters 22 and 33) seem likely can-
didates for such evolutionarily modified features; the capital shaft ridge (22)
also may be similarly modified in Chenonetta, which shows a reversal in this
character (Livezey, 1986).

Inclusion of Cnemiornis in the analysis indicated that, of the three
alternative sequences of the extant taxa Dendrocygna, Thalassornis, and

22 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Anserinae (Fig. 6), placement of the Anserinae as the sister-group of the rest
of the Anatidae is most parsimonious (Fig. 5). The relatively poor resolution
of the phylogenetic sequence of the Anserinae and Thalassominae was
indicated previously by Livezey (1986) and also by an analysis of the
Miocene fossil Mionetta blanchardi (Livezey and Martin, 1988). The impor-
tance of fossils in the (cladistic) inference of phylogenetic relationships of
extant taxa was demonstrated by Panchen and Smithson (1987) and Gauthier
et al . ( 1 988). In practice, however, this advantage must be weighed against the
disadvantages associated with incomplete data, worn material, uncertain
association, and unknown sex of many specimens of fossil and subfossil taxa.

The phylogenetic hypotheses proposed here add to the growing evidence
of the diversity and probable origin of the Anseriformes in the Southern
Hemisphere (Livezey, 1986). The very early divergence of Cnemiornis is
concordant with the numerous flightlessness-related autapomorphies in the
genus, in that it provides a much greater period of time for the accumulation
of these evolutionary novelties. A parallel example of morphologically
radical flightlessness in an ancient carinate lineage is the gruiform Aptomis
(also endemic to New Zealand); formeriy thought to be a rail (Oliver, 1945),
Aptornis instead probably represents a separate family related to the Rallidae
(cf. Olson, 1975).

The finding ihdXEuryanas is nota member of the Anatinae underscores the
diversity of the more primitive "proto-ducks" in the Southern Hemisphere in
the past. Moreover, Euryanas, the extant Thalassornis of Africa and Sticton-
etta of Australia (Livezey, 1986), and several Miocene forms from the
Northern Hemisphere (Livezey and Martin, 1988) indicate that there was a
more widespread radiation of these "duck-Uke" anseriforms in the late
Tertiary.

ACKNOWLEDGEMENTS

This study was supported by National Science Foundation (USA) grant
BSR-85 16623. I am grateful for the arrangements and kind hospitality
afforded me by: J. Darby, Otago Museum; G. Tunnicliffe, Canterbury
Museum; J. A. Bartle and N. H. S. Hyde, National Museum of New Zealand;
B. Gill, Auckland Museum; curatorial staffs of the U. S. National Museum
and the Museum of Natural History, University of Kansas; and C. Walker,
British Museum (Natural History). P. S. Humphrey, D. Siegel-Causey, and
two anonymous reviewers commented on the manuscript, and S. L. Olson
provided useful criticisms of an earlier version of this paper; K. Corbin and
M. Schmalz typed its several drafts.

SUBFOSSIL ANSERIFORMES 23

SUMMARY

The phylogenetic relationships of several endemic subfossilAnseriformes
of New Zealand — Cnemiornisspp.,Euryanasfinschi,andCygnussumneren-
sis — are re-examined using the osteological characters analysed in an earlier
study of anseriform systematics (Livezey, 1986). Flightless Cnemiornis,
traditionally considered to be a "goose" and closely related to the extant
Australian genus Cereopsis, is shown to be a very primitive anseriform
representing a branch shortly after that ofAnseranas of Australia. Euryanas
finschi is found to be a moderately derived "proto-duck," most probably
representing the sister-group to the clade including Tadominae and Anatinae.
Cygnus sumnerensis is confirmed to be a true swan (Cygnini), more derived
than Coscoroba but less so than 0 /or. A revised classification is presented and
selected biogcographic, analytical, and evolutionary implications are dis-
cussed.

LITERATURE CITED

Brodkorb, P. 1964. Catalogue of Fossil Birds: Part 2 (Anseriformes through Galliformes). Bull.
Florida State Mus. (Biol. Sci.) 8:195-335.

Cracraft, J. 1980. Phylogenetic theory and methodology in avian paleontology: a critical
appraisal. Conlr. Sci. Natur. Hist. Mus. Los Angeles County 330:9-16.

Davies, S. J. J. F. and H. J. Frith. 1964. Some comments on the taxonomic position of the Magpie
Goose Anseranas semipalmala (Latham). Emu 63:265-272.

Dawson, E. W. 1958. Re-discoveries of the New Zealand subfossil birds named by H. O. Forbes.
Ibis 100:232-237.

Delacour, J. 1954. The Waterfowl of the World. Vol. 1. Country Life Ltd., London.

Delacour, J. 1959. The Waterfowl of the World. Vol. 3. Country Life Ltd., London.

Delacour, J. 1964. Corrections and additions. In J. Delacour (ed.). The Waterfowl of the World.
Vol. 4, pp. 327-354. Country Life, London.

Delacour, L and E. Mayr. 1945. The family Anatidae. WQson Bull. 57:1-55.

Forbes, H. O. 1890a. [Announcement of paper given on 3 October 1889 at meeting of
Philosophical Institute of Canterbury, New Zealand.] Nature 41:209.

Forbes, H. O. 1890b. New extinct swan in New Zealand. Ibis 32:264-265.

Forbes, H. O. 1891. Preliminary notice of additions to the extinct avifauna of New Zealand.
Trans. New Zealand Inst. 24:185-189.

Forbes, H. O. 1892a. On a recent discovery of the remains of extinct birds in New Zealand.
Nature 45:416-418.

Forbes, H. O. 1892b. On a recent discovery of the remains of extinct birds in New Zealand.
Science 19:163-165.

Gauthier, J., A. G. Kluge, and T Rowe. 1988. Amniote phylogeny and the importance of fossils.
Qadistics 4:105-209.

Hamilton, A. 1 892. On the fissures and caves at the Castle Rocks, Southland; with a description
of the remains of the existing and extinct birds found in them. Trans. New Zealand Inst.
25:88-106.

24 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY

Hector, J. 1873a. On Cnemiornis calcitrans, showing its affinity to the Natatores. Proc. Zool.
Soc. London 1873:763-76L

Hector, J. 1873b. On Cnemiornis calcitrans, Owen, showing its affinity to the lamellirostrate
Natatores. Trans. New Zealand Inst. 6:76-84.

Hector, J. 1874. [Correction of table in Hector 1873a:763]. Proc. Zool. Soc. London 1874:307.

Howard, H. 1929. The avifauna of Emeryville shellmound. Univ. Calif. Publ. Zool. 32:301-394.

Howard, H. 1964. FossQ Anserifonnes. In J. Delacour (ed.). The Waterfowl of the World. Vol.
4, pp. 233-326. Country Life, London.

Johnsgard, P. A. 1961a. The taxonomy of the Anatidae — a behavioral analysis. Ibis 103a:71-85.

Johnsgard, P. A. 1961b. Tracheal anatomy of the Anatidae and its taxonomic significance.

Wadfowl 12:58-^9.
Lambrecht, K. 1933. Handbuch der Palaeomithologie. Gebruder Bomtraeger, Berlin.

Livezey, B. C. 1986. A phylogenetic analysis of Recent anseriform genera using morphological
characters. Auk 103:737-754.

Livezey, B. C. and L. D. Martin. 1988. The systematic position of the Miocene anatid AnasYl]
Wanc/wrdj Milne-Edwards. L Vert. Paleontol. 8:196-211.

Lydekker, R. 1891. Catalogue of the Fossil Birds in the British Museum (Natural History). Brit.
Mus. (Natur. Hist.), London.

MUler, A. H. 1937. Structural modifications in the Hawaiian Goose (Nesochen sandvicensis):
A study in adaptive evolution. Univ. Calif. Publ. Zool. 42: 1-80.

Oliver, W. R. B. 1930. New Zealand Birds, 1st ed. A. H. & A. W. Reed, Wellington.

Oliver, W. R. B. 1945. Avian evolution in New Zealand and Australia. Emu 45:119-152.

Oliver, W. R. B. 1955. New Zealand Birds, 2nd ed. A. H. & A. W. Reed. Wellington.

Olson, S. L 1975. A review of the extinct rails of the New Zealand region (Aves: RaUidae). Nat.
Mus. New Zealand Rec. 1 :63-79.

Olson, S. L. and H. F. James. 1982. Prodromus of the fossil avifauna of the Hawaiian Islands.
Smithsonian Contr. Zool. No. 365:1-59.

Olson, S. L. and A. Wetmore. 1976. Preliminary diagnoses of two extraordinary new genera of
birds from Pleistocene deposits in the Hawaiian Islands. Proc. Biol. Soc. Washington
89:247-258.

Owen, R. 1866. On Dinornis (Part X): containing a description of part of the skeleton of a
flightless bird indicative of a new genus and species {Cnemiornis calcitrans, Ow.). Trans.
Zool. Soc. London 5:395^04.

Owen, R. 1 875. On Dinornis (Part XX): containing a restoration of the skeleton of Cnemiornis
calcitrans, Ow., with remarks on its affinities in the lamellirostral group. Trans. Zool. Soc.
London 9:253-292.

Owen, R. 1879. Memoirs on the Extinct Wingless Birds of New Zealand with an Appendix on
Those of England, Australia, Newfoundland, Mauritius, and Rodriquez. 2 vols. John van
Voorst, London.

Panchen, A. L. andT. R. Smithson. 1987. Character diagnosis, fossils and the origin of tetrapods.
Biol. Rev. 62:341-438.

Schoch, R. M. 1986. Phylogeny Reconstruction in Paleontology. Van Nostrand Reinhold, New
York.

Sclater, P. L. 1890. New extinct swan in New Zealand. Ibis 32:264-265.

Shufeldt, R. W. 1913. On the comparative osteology of Cereopsis novae-hollandiae. Emu
12:209-237.

Stejneger, L. 1885. Order VHI. — Chenomorphae. In J. S. Kingsley (ed.). The Standard Natural

SUBFOSSIL ANSERIFORMES 25

History. Vol. IV, pp. 132-157. S. E. Casino, Boston.

Swofford, D. L. 1985. PAUP: Phylogenetic Analysis Using Parsimony. Version 2.3. Illinois
Natur. Hist. Surv. Progr. Man., Urbana.

Van Beneden, P. J. 1875. Un oiseau fossUe nouveau des cavemes de la Nouvelle Zelande. Annal.
Soc. Geol. Belgique 2:123-130.

Van Beneden, P. J. 1876. A new fossQ bird. Anas finschi, from the Eamscleugh Caves, Olago,
New Zealand. Trans. New Zealand Inst. 9:599-602.

Verheyen, R. 1953. Bijdrage lot de Osteologie en de Systematiek der Anseriformes. Gerfaut
43:373-456.

Wetmore, A. 1943. An extinct goose from the island of Hawaii. Condor 45:146-148.

Wiley, E. O. 1981. Phylogenetics: The Theory and Practice of Phylogenetic Systematics. John
Wiley and Sons, New York.

Woolfenden, G. E. 1961 . Postcranial osteology of the waterfowl. BuU. Florida State Mus. (Biol.
Sci.) 6:1-129.

Worthy, T. H. 1988. Loss of flight ability in the extinct New Zealand duck Euryanasfinschi. J.
Zool. (London) 215:619-628.

Umversity of Kansas Publications

MUSEUM OF NATURAL HISTORY

The University of Kansas Publications, Museum of Natural
History, begirming with volume 1 in 1946, was discontinued with
volume 20 in 197 1 . Shorter research papers formerly published in
the above series are now published as Occasional Papers, Museum
of Natural History. The Miscellaneous Publications, Museum of
Natural History, began with number 1 in 1946. Longer research
papers are published in that series. Monographs of the Museum of
Natural History was initiated in 1970. Authors should contact the
managing editor regarding style and submission procedures before
manuscript submission. All manuscripts are subject to critical
review by intra- and extramural specialists; final acceptance is at
the discretion of the Director. The Occasional Papers and Miscel-
laneous Publications are produced by a Macintosh computer, Las-
erWriter, and the software Microsoft® Word and PageMaker®.
Conversion to current format by E. O. Wiley.

Institutional libraries interested in exchanging publications
may obtain the Occasional Papers and Miscellaneous Publications
by addressing the Exchange Librarian, University of Kansas Li-
brary, Lawrence, Kansas 66045. Individuals may purchase sepa-
rate numbers of all series. Prices may be obtained upon request ad-
dressed to Publications Secretary, Museum of Natural History,
University of Kansas, Lawrence, Kansas 66045 - 2454 U.S.A.

Editor: Robert D. Holt

Managing Editor: Joseph T. Collins

Design and Typesetting: K. A. Shaw and Joseph T. Collins

PRINTED BY

UNIVERSITY OF KANSAS PRINTING SERVICE

LAWRENCE. KANSAS

DATE DUE

'MAR 31 1996

OEMCO, INC. 38-2931

3 2044 093 361 707